In this study, we demonstrated noninvasive, quantitative, and tomographic imaging of tumor xenografts and metastases in mouse models using a combination of small-animal PET and small-animal CT. We determined the feasibility of coregistering the PET and CT images in xenograft models, explored the possibility of using 18F-FDG PET/CT to detect pulmonary melanoma metastases, and evaluated the detection of systemic metastases of a reporter gene–expressing melanoma cell line with 18F-FHBG PET/CT and BLI.
Our method was based on the use of fiducial markers for fusion of the PET images with the CT images. Because the alignment error was about 5 times lower than the resolution of the small-animal PET scanner, we consider this method sufficiently accurate. We did not perform any partial-volume or attenuation correction, because this was beyond the scope of the study. The reported values are therefore underestimations for lesions smaller than about 4 mm in diameter.
The tumor xenografts could be visualized with 18
F-FDG PET and CT. The N2a xenografts showed increasing uptake with time, whereas the C6 xenografts showed no differences in mean or maximum uptake over time. This finding was due to a partial-volume effect in the N2a xenografts, because the volume in the early scans was around 45 mm3
(diameter, ~4.5 mm). In the C6 xenografts, the increase in mean uptake was less than the increase in the maximum pixel value because of necrosis in the tumor periphery. The fact that this necrosis was not seen on the CT images in any of the tissue windows we examined was due to the poor soft-tissue contrast of small-animal CT. This finding highlights the need for contrast agents to characterize tumors in small-animal CT. These agents can be of the classic iodine-based type, can be tissue-specific (12
), or can come from another imaging modality, as in our approach using 18
To study metastasis, we first used an established model of lung metastasis (7
) to take advantage of the contrast between tumoral soft tissue and the air in the lungs. We were successful in detecting pulmonary metastases with CT and 18
F-FDG PET in 4 of 5 SCID mice 45 d after intravenous injection of A375M-Fluc melanoma cells. These mice did represent an advanced stage of disease, with lesions ranging in size from 7.5 to 58 mm3
. High background uptake of 18
F-FDG in heart, muscle, and brown fat did interfere with lesion delineation in 2 mice, and because pulmonary metastases of B16-F0 melanoma cells did not show increased 18
F-FDG uptake, we further focused on reporter gene–based metastasis detection.
We used melanoma cells expressing the trimodality fusion reporter gene hRL-mRFP-tTK after lentiviral transduction (A375M-3F). We showed that reporter gene activity was still present after 12 cell culture generations, indicating that there is no significant silencing of reporter gene expression that might influence cell detection. Ten meta-static sites in total were detected by 18
F-FHBG PET, with a mean uptake of 3.3 ± 1.3 %ID/g. The CT images showed changes in the lung and liver but not in the bone lesions. Earlier reports showed the presence of osteolytic lesions caused by A375 cells (17
), but this is not the case for the derived cell line A375M, which was selected for its capacity to form lung metastases (19
). Despite a higher uptake of 18
F-FDG in the A375M-3F cells, 18
F-FHBG was superior to 18
F-FDG because of its much lower background uptake in the thorax and the head. Tumor-to-background contrast showed a significant increase of at least 0.27 for 18
F-FHBG in the most relevant thoracic organs. This better performance of 18
F-FHBG than of 18
F-FDG in the thorax was anticipated because 18
F-FHBG is specifically designed to be retained only in cells expressing HSV1-tk, a viral enzyme that in these mice is present only in the tumor cells. Furthermore, 18
F-FHBG has a favorable pharmacokinetic profile, with rapid washout from normal tissues. In contrast, 18
F-FDG uptake is related to glucose metabolism and is determined primarily by the number of active glucose transporters on the cell membrane and by hexokinase activity, and these are not restricted to tumor cells but also occur in brain, heart, brown fat, and liver. In intestines, the nonspecific physiologic hepatobiliary clearance of 18
F-FHBG causes higher activity levels that are responsible for the limited detection of metastatic lesions in the peritoneal cavity, the lumbar spine, and pelvic bones. Novel molecules with better imaging characteristics, such as 2′-fluoro-2′-deoxyarabino-furanosyl-5-ethyluracil (21
), are currently under investigation in our laboratory, and they might improve the detection of metastasis because they show a higher absolute uptake and lower gastrointestinal activity than does 18
F-FHBG. Using a rabbit model of VX2-carcinoma pulmonary metastasis, Kondo et al. (22
) showed that the performance of 18
F-FDG PET was significantly greater for lesions larger than 4 mm than for smaller ones. Recently, oncologic imaging of mice has been reported in clinical combined PET/CT systems (23
), but the low resolution of the system (6 mm) interferes with accurate quantification of metastatic lesions.
The BLI signal intensity is depth-dependent, resulting in a higher signal for superficial lesions than for deeper lesions that showed a comparable 18
F-FHBG uptake, confirming our previous observations (9
). BLI has the advantage of a relatively low cost and high throughput capability, making the depth dependence of the signal the major disadvantage in small animals. The other major limitation of BLI is that the planar nature of the images impedes accurate 3-dimensional localization of the signal. Recent studies have used BLI for evaluating metastases noninvasively (24
), for evaluating lymphoma therapy (25
), for studying the influence of differential gene expression on the pattern of metastasis formation (26
), and for studying the importance of cellular receptors for metastasis development (27
). The development of tomographic optical systems is progressing (28
). These systems are using novel hardware that allows the acquisition of bioluminescence images from multiple angles and the use of reconstruction algorithms that apply models of photon transport in tissue to localize and quantify photon sources, but tomographic optical systems still need substantial validation regarding localization and quantification of the luminescent signal. The low cost, high sensitivity, and high throughput of planar BLI also make it suited as a routine monitoring tool that, when metastases are detected, can be complemented by 18
F-FHBG small-animal PET/CT for anatomic localization and quantitation. A slightly different strategy was used by Kang et al. (29
). They used BLI to detect MDA-MB-231 breast cancer bone metastases followed by PET reporter gene imaging of HSV1-tk expression under the control of a TGF-β response element. Images produced by the latter represent signal transduction through the TGF-β/Smad pathway and can show in vivo that these cells are engaged in Smad-dependent transcription while growing as bone metastases. Thus, PET and BLI can be used to detect 2 different molecular signals from the same metastasis.
We have demonstrated that multimodality imaging with PET, CT, and BLI is a powerful tool in the study of meta-static mouse models. Because of the low contrast of 18F-FDG, even in a cell line that has a high 18F-FDG uptake, PET reporter gene strategies are superior to 18F-FDG imaging except for abdominal metastases. On the basis of our results, we suggest the following imaging strategy: frequent BLI scanning should be used initially to monitor the animals while the metastases are below the detection thresholds of PET and CT. When BLI shows lesions that have reached the detection thresholds of PET or CT, combined PET/CT scans can than be used for detection (PET and CT), quantitation (PET), anatomic localization (CT), or density and volume measurement (CT).
A major limitation of the current study is the lack of a histologic gold standard for assessing tumor metastases. For identifying lesions as metastases, this lack was partially compensated for by the double detection of lesions on PET and BLI and by the in vitro analysis of tissue extracts. However, there might have been lesions that did not reach the threshold of detection by either imaging method, resulting in false-negative findings. This study, however, was designed to show the feasibility of using multimodality PET/CT and BLI to detect metastases. Future work will have to address the issue of sensitivity and determine the detection thresholds for the different modalities. For instance, scan times longer than 10 min for PET might increase sensitivity. Histologic standards will be needed, and the monomeric red fluorescent protein moiety of the multimodality reporter gene will be useful for detecting and identifying fluorescent metastatic cells in tissue specimens. The monomeric red fluorescent protein moiety is also important for fluorescence-activated cell sorting, rather than for in vivo fluorescent imaging, in which the high background and the weak signal generate a signal-to-noise ratio that is much lower than for BLI (30